PackDiT: Joint Human Motion and Text Generation via Mutual Prompting
This work addresses the need for aligning diverse modalities in human motion generation, enabling tasks like motion-to-text and joint generation, which is incremental by extending diffusion models to bidirectional capabilities.
The paper tackles the problem of bidirectional generation between human motion and text, introducing PackDiT, a diffusion-based model that achieves state-of-the-art text-to-motion performance with an FID score of 0.106 and competitive motion-to-text results.
Human motion generation has advanced markedly with the advent of diffusion models. Most recent studies have concentrated on generating motion sequences based on text prompts, commonly referred to as text-to-motion generation. However, the bidirectional generation of motion and text, enabling tasks such as motion-to-text alongside text-to-motion, has been largely unexplored. This capability is essential for aligning diverse modalities and supports unconditional generation. In this paper, we introduce PackDiT, the first diffusion-based generative model capable of performing various tasks simultaneously, including motion generation, motion prediction, text generation, text-to-motion, motion-to-text, and joint motion-text generation. Our core innovation leverages mutual blocks to integrate multiple diffusion transformers (DiTs) across different modalities seamlessly. We train PackDiT on the HumanML3D dataset, achieving state-of-the-art text-to-motion performance with an FID score of 0.106, along with superior results in motion prediction and in-between tasks. Our experiments further demonstrate that diffusion models are effective for motion-to-text generation, achieving performance comparable to that of autoregressive models.